Fine-grained classification method and device and computer readable storage medium

Abstract

The invention relates to artificial intelligence, and discloses a fine-grained classification method, which comprises the following steps of creating an initial model, acquiring an original image, andpreprocessing the original image to form training data for training the initial model; acquiring loss data corresponding to the original image based on the initial model and the training data; performing random gradient descent back propagation based on the loss data until the initial model is converged within a preset range to form classification model training; and performing classification prediction on the to-be-processed image through the classification model. The invention also relates to a blockchain technology, and the original image is stored in the blockchain. According to the method, the local feature region can be adaptively selected, and the classification prediction accuracy is improved.

Description

technical field [0001] The present invention relates to artificial intelligence, in particular to a fine-grained classification method, device, electronic equipment and computer-readable storage medium. Background technique [0002] Fine-grained classification mainly refers to dividing the basic categories into finer subcategories, such as distinguishing the types of birds and the styles of cars. Currently, it has a wide range of business needs and application scenarios in the industry and in real life. [0003] Currently, fine-grained classification methods mainly include two categories. One is to classify by locating local feature regions and extracting these image features that can distinguish between different categories of regions. However, since there is no location labeling information of local key regions, most algorithms use pre-set size windows to select local feature regions. It can be seen that this method cannot automatically adapt to the size of the feature a...

AI Technical Summary

Problems solved by technology

However, the previous algorithms based on this idea cannot achieve end-to-end training, and the training process is repeated.

It cannot be automated. In addition, during the entire training process, continuous human intervention is required, the training time is long, and the expected training results are uncertain

Images